Computer Software in Science and Mathematics Computation offers a new means of describing and investigating scientific and mathematical systems. Simulation by computer may be the only way to predict how certain complicated systems evolve by Stephen Wolfram cientific laws give algorithms, or given the equations that describe the tion is specified by a set of numbers procedures, for determining how motion of electrons in an arbitrary mag­ stored in a computer. The computer S systems behave. The computer netic field, it is possible to derive a sim­ program applies an algorithm that sim­ program is a medium in which the algo­ ple mathematical formula that gives the ulates the motion of the electron by rithms can be expressed and applied. trajectory of an electron in a uniform changing the numbers representing its Physical objects and mathematical magnetic field (one whose strength is position at successive times. Computers structures can be represented as num­ the same at all positions). For more are now fast enough for the simulations bers and symbols in a computer, and a complicated magnetic fields, however, to be carried out quickly, and so it is program can be written to manipulate there is no such simple mathematical practical to explore a large number them according to the algorithms. When formula. The equations of motion still of cases. The investigator can interact the computer program is executed, it yield an algorithm from which the directly with the computer, modifying causes the numbers and symbols to be trajectory of an electron can be deter­ various aspects of a phenomenon as new modified in the way specified by the sci­ mined. In principle the trajectory could results are obtained. The usual cycle of entific laws. It thereby allows the conse­ be worked out by hand, but in prac­ the scientific method, in which hypothe­ quences of the laws to be deduced. tice only a computer can go through the ses are formulated and then tested, can Executing a computer program is large number of steps necessary to ob­ be followed much faster with the aid of much like performing an experiment. tain accurate results. the computer. Unlike the physical objects in a conven­ A computer program that embodies tional experiment, however, the objects the laws of motion for an electron in a omputer experiments are not limit­ in a computer experiment are not bound magnetic field can be used to perform ed to processes that occur in nature. by the laws of nature. Instead they fol­ computer experiments. Such experi­ ForC example, a computer program can low the laws embodied in the computer ments are more flexible than conven­ describe the motion of magnetic mono­ program, which can be of any consistent tional laboratory experiments. For ex­ poles in magnetic fields, even though form. Computation thus extends the ample, a laboratory experiment could magnetic monopoles have not been de­ realm of experimental science: it allows readily be devised to study the trajecto­ tected in physical experiments. More­ experiments to be performeq in a hypo­ ry of an electron moving under the influ­ over, the program can be modified to thetical universe. Computation also ex­ ence of the magnetic field in a television embody various alternative laws for the tends theoretical science. Scientific laws tube. No laboratory experiment, how­ motion of magnetic monopoles. Once have conventionally been constructed in ever, could reproduce the conditions en­ again, when the program is executed, terms of a particular set of mathemati­ countered by an electron moving in the the consequences of the hypothetical cal functions and constructs, and they magnetic field surrounding a neutron laws can be determined. The comput­ have often been developed as much star. The computer program can be ap­ er thus enables the investigator to ex­ for their mathematical simplicity as for plied in both cases. periment with a range of hypothetical their capacity to model the salient fea­ The magnetic field under investiga- natural laws. tures of a phenomenon. A scientific law specified by an algorithm, however, can have any consistent form. The study of many complex systems, which have re­ COMPUTER SIMULATION has made it practical to consider many new kinds of models for sisted analysis by traditional mathemat­ natural phenomena. Here the stages in the formation of a snowflake are generated by a com­ ical methods, is consequently being puter program that embodies a model called a cellular automaton. According to the model, the made possible through computer exper­ plane is divided into a lattice of small, regular hexagonal cells. Each cell is assigned the value 0, iments and computer models. Compu­ which corresponds to water vapor (black), or the value 1, which corresponds to ice (color). Be­ tation is emerging as a major new ap­ ginning with a single red cell in the center of the illustration, the simulated snowflake grows in proach to science, supplementing the a series of steps. At each step the subsequent value of any cell on the boundary of the snow­ long-standing methodologies of theory flake depends on the total value of the six cells that surround it. If the total value is an odd num­ ber, the cell becomes ice and takes on the value 1; otherwise the cell remains vapor and keeps and experiment. the value O. The successive layers of ice formed in this way are shown as a sequence of colors, There are many scientific calcula­ ranging from red to blue every time the number of layers doubles. The calculation required for tions, of course, that can be done by con­ each cell is simple, but for the pattern shown more than 10,000 calculations were needed. The ventional mathematical means, without only practical way to generate the pattern is by computer simulation. The illustration was made the aid of the computer. For example, with the aid of a program written by Norman H. Packard of the Institute for Advanced Study. 188 © 1984 SCIENTIFIC AMERICAN, INC This content downloaded from 130.126.162.126 on Tue, 08 Oct 2019 17:21:16 UTC All use subject to https://about.jstor.org/terms The computer can also be used to trons traveling through the magnetic where a is a number that can range be­ study the properties of abstract math­ fields in a circular particle accelerator. tween 0 and 4. The formula gives an ematical systems. Mathematical exper­ The transverse displacement of an elec­ algorithm from which the sequence of iments carried out by computer can of­ tron as it passes a point on one of its values for the electron's displacement ten suggest conjectures that are sub­ revolutions around the accelerator ring can be worked out. sequently established by conventional is given by some fraction x between 0 A few trials show how the properties mathematical proof. Consider a math­ and 1. The value of the fraction corre­ of the sequence depend on the value of ematical system that can be introduced sponding to the electron's displacement a. If a is equal to 2 and the initial value to model the path of a beam of elec- on the next revolution is then ax(1 - x), of x is equal to .8, the next value of x, © 1984 SCIENTIFIC AMERICAN, INC This content downloaded from 130.126.162.126 on Tue, 08 Oct 2019 17:21:16 UTC All use subject to https://about.jstor.org/terms PHYSICAL PROCESS ALGORITHMIC DESCRIPTION COMPUTER EXPERIMENT (40 STEPS) TRIAL 1 TRIAL2 EXACT SOLUTION � l' >­ u. I­ �--------------�--------------�Ooo :::::i TRIAL 100 a:....J jjJ UJ« « co o ��::::lI- a: Z Cl. -15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15 POSITION POSITION NUMERICAL APPROXIMATION TIME 4 ••• TIME 14 .,. TIME 40 >­ tI­ :::::i jjJ « CO o a: Cl. -15 -10 -5 0 5 10 15 -15 -10 10 15 -15 -10 -5 0 5 15 POSITION POSITION POSITION ALGEBRAIC APPROXIMATION (40 STEPS) ORDER 0 ORDER1 ORDER2 4�" � �40" (A2X')2x40' Y � «>-' 1' , \ a: ....J Cl.jjJ13t: -15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15 -15 -10 -5 5 10 15 POSITION POSITION POSITION MATHEMATICAL AND COMPUTATIONAL METHODS are constructed for the distribution, and the equation is simple enough applied in various ways in the study of random walks. A random walk for an exact solution to be given. For most differential equations, is a model for such physical processes as the Brownian motion of a however, no such exact solution can be obtained, and approximations small particle suspended in a liquid. The particle undergoes random must be made. In numerical approximations the smooth variation of deflections as it is bombarded by the molecules in the liquid; its path quantities in the differential equation is approximated by a large can thus be described as a sequence of steps, each taken in a random number of small increments. The results shown in the diagram were direction. The most direct way to deduce the consequences of the obtained by a computer program in which the spatial and temporal model is by a computer experiment. Many random walks are simulat­ increments were small fractions of the lengths and times for individ­ ed on a computer and their average properties are measured. The dia­ nal steps in the random walk. Algebraic approximations to the differ­ gram shows a histogram in which the height of each bin records the ential equation are found as a series of algebraic terms. The diagram number of simulated random walks that were found to have reached shows the first three terms in such a series. The contribution of each a particular range of positions after a certain time.